Training Computers to Understand the Human Brain
In keeping with yesterdays post ( Using brain imaging to predict your behaviour ) researchers from Tokyo Institute of Technology, collaborating with co-workers in the USA, UK and Italy, have completed a study that used fMRI datasets to train a computer to predict categories of images.
To do this, they had volunteers come in and look at pictures of different animals or tools. They were asked to categorise these silently, while undergoing a fMRI scan. These scans were then analysed for patterns that would separate these two categories (animals and tools).
After getting these algorithms, and in some way “training” the computer - the computer was able to correctly identify the scans with 90% accuracy. So - the computer was able to look at the volunteers’ scans, analyse them, and predict what category they were looking at.